Now you meet an agent by running one, which is the fastest way to understand what all the fuss is about. You will configure a persistent agent with Hermes, connect it to WhatsApp, and watch it take action. Crucially, this is configuration and skills, not code, so the whole room can operate a real agent without building one.

Here is what "operate, not build" means and why it is the right call for this week. Building an agent from scratch (the loop, the orchestration, the plumbing) is what the Engineering Bootcamp teaches. You do not need to write that to understand it; you need to run one, direct it, and watch it behave. So you will set up the agent's instructions and give it the tools and skills it needs, then send it a real task and see it decide and act. Watching an agent receive a goal in WhatsApp and then take steps toward it, on its own, is the moment the previous lesson's distinction stops being abstract.

Pay attention, as it runs, to where it is deciding versus following. That is the line from the last lesson, now visible in something you are operating. When the agent chooses to use one tool over another, or takes an extra step you did not explicitly script, that is the "agent" part, the model directing its own process toward the goal you set.

You will leave with a live agent wired into a real channel and, more importantly, a felt understanding of what an agent is that no amount of reading could give you.